We all sit around and accept the changes that are brought upon us in life, but do you ever wonder what brings these dramatic changes around? Did it ever cross your mind that there was someone out there who decided that a T-Shirt would be more comfortable to wear than a dress shirt? Or at one point in time the standard paper size was decided to be 8.5" x 11"? Things like this we normally don't pay much attention to, but when it comes to the computer hardware industry it seems like we must otherwise we end up left in the dust by the rapidly changing industry.

Back in 1997, with Intel's release of the i430TX chipset one of the most highly boasted features of the new chipset standard was its support for the Ultra ATA/33 hard drive interface standard. Ultra ATA/33, by definition, allowed for burst transfer rates of up to 33.3MB/s for compliant EIDE devices over the PCI bus. Ultra ATA/33 was, at the time, the latest attempt at a low-cost competitor to the high-end SCSI standard for storage devices. The reason for the move to Ultra ATA/33, which was an effective doubling of the previous burst transfer rate standard for EIDE devices (DMA Mode 2 - PIO Mode 4) was basically because of the internal improvements in EIDE hard drives, making the drives reach a point where they could retrieve data internally faster than they could send it to the host controller. The situation provided a bit of a dilemma, as any case where a performance bottleneck is present would, in this case, the easiest solution came in the form of the Ultra ATA/33 standard which doubled burst transfer rates and bought the industry a couple more years until the performance bar needed to be lifted once again.

For those of you that were in to computer hardware when the TX chipset became popular it's quite difficult to remember exactly when Ultra ATA/33 took off, as it was a highly criticized "feature" due to its relatively small performance improvement over previous standards. Today, if you look at any EIDE hard drive, chances are you won't find anything that isn't Ultra ATA/33 compliant, isn't it funny how changes come to be?

Just as the industry reached that limitation in 1997, the time for the next "big" jump in hard drive interface standards is upon us, let's say hello to Ultra ATA/66.

Here's the problem, you have this data stored on your hard drive and your computer needs to store it in system memory so you can manipulate it. This problem is something that we all face every day, whenever you open a document on your hard drive the disk reads the file, places it in its own buffer (memory) and copies it to your system RAM via your hard disk controller host's bus. The transfer protocol the host's bus uses states the theoretical maximum for performance out of the drive itself, so regardless of how fast your hard drive can read data off of its platter, the transfer protocol is the final decision maker as to what kind of performance you can achieve. Confused?

Let's see if we can simplify things, assume we have a hard drive that can read data off of its platter (data is stored on circular plates known as platters) at a rate of 2MB per second, and let's also assume that the drive has a 1MB buffer. In a single second, the drive is ready to transfer 1MB of data from the buffer to the computer's system memory and provided that the host bus is capable of providing transfer rates of up to 1MB per second, your hard drive will be able to operate at its potential without being restricted in any manner.

Hard drives cannot transfer at their theoretical maximum 100% of the time, in fact, very rarely do they achieve the maximum performance set by their transfer protocols, instead they achieve their theoretical maximum levels of performance in bursts, which is where the term burst transfer rate comes from. This is different from sustained transfer rate which is often much lower than the theoretical maximum of the drive, and is generally the performance you can expect from your hard drive.

Let's take a look at the same hard drive from before, and let's assume that the host bus is capable of providing transfer rates of up to 512KB (1/2MB) per second. While your hard drive may be ready to transfer the 1MB of data it just read from its platter to its buffer after a second, the host bus can only allow for a maximum of half of that data, or 512KB, to be transferred from the buffer to the computer's system memory in a single second. Here's where the problem arises, since the example hard drive only has a 1MB buffer, while waiting for the first 512KB of data to transfer down the bus it can only read another 512KB into its buffer since the bus can only transfer at 512KB per second. It's like having a 100-gallon water tank but only having a pump capable of moving 1 gallon per hour, you have the potential to do much more, but an external force is limiting you.

Copyright 1999 Western Digital

This was the case with the old DMA Mode 2 (or PIO Mode 4) transfer protocol standard whose 16.6MB/s burst transfer rate was quickly becoming a bottleneck for newer hard drives. At that point came the need for Ultra ATA/33 that effectively doubled the maximum burst transfer rate to 33.3MB/s. Two years later, it seems as if the advancements in hard disk storage technology have brought us to that limit once again, so break out the calculators, as we're about to double the theoretical maximum once again, Ultra ATA/66 is upon us.

Ultra ATA/66 (you'll also hear it referred to as Ultra DMA/66 or Fast ATA-2) allows for a theoretical maximum burst transfer rate of 66.6MB/s, double the maximum of Ultra ATA/33's 33.3MB/s. Another change brought about during the introduction of the Ultra ATA/33 specification was the idea of Cyclical Redundancy Check (CRC) which is something that you've all probably heard of in one form or another at sometime during your computing experience. Basically, before any burst transfers take place both the hard drive and host will calculate what should make it through the transfer and then afterwards the host sends a signal back to the hard drive telling it what actually made it over during the transfer. If the two numbers don't match up then the process is automatically repeated until it is completed successfully. Ultra ATA/66 brings the same CRC feature that Ultra ATA/33 brought to the table, couples it with its 66.6MB/s theoretical burst transfer rate, and adds yet another new feature to the growing list: improved data integrity.

While CRC increased the integrity of transferred data, after approaching burst speeds of 33.3MB/s we begin to truly stress the capabilities of conventional 40-pin IDE cables, and there is a considerable increase in signal "leakage" at the more aggressive timings. The 44.4MB/s mark (Ultra ATA mode 3) is the first marker of unreliability when using conventional cables during burst transfers, what happens is that the data loses its integrity between the hard drive and host controller due to increased timings that the cables were never intended to deal with. At the same time, a standard that required a completely new cable design (such as something similar to what SCSI devices use to achieve their 40MB/s+ burst transfer rates) would never be pursued by the market since that would mean zero backwards compatibility, the solution? Reuse the same cables with a new twist, twice as many conductors.

The Ultra ATA/66 specification calls for the same 40-pin IDE cables that all IDE/EIDE hard drives have used for years now however, in order to take advantage of the specification's higher burst transfer rates the cable specification had to be modified to take into account the more aggressive timing signals that would be transferred over the cables. The result was the same 40-pin IDE cable with twice as many ground lines present in the cable to act as shields between the lines that actually carry live signals back and forth. By doing this we maintain backwards compatibility with all older IDE/EIDE hard drives and ensure data integrity during burst transfers exceeding 33.3MB/s.

Copyright 1999 Western Digital

What do you need to take advantage of Ultra ATA/66? There are 4 basic requirements:

• Ultra ATA/66 support in your motherboard's chipset, currently the only available chipset that supports Ultra ATA/66 is the VIA Apollo Pro+. The Apollo Pro+ or any VIA chipsets for that matter can take advantage of Ultra ATA/66 as long as they make use of VIA's VT82C686A Super South Bridge Controller, which is a bit more expensive than the VT82C596A Mobile South Bridge Controller that most motherboard manufacturers opt to use. AnandTech has yet to see any Slot-1 Apollo Pro+ boards in the lab that use the updated South Bridge Controller, however AnandTech's MVP4 sample did have the controller and that is what was used in the tests.  Intel's upcoming Camino chipset should support Ultra ATA/66.
• Operating System support for DMA transfers via a DMA device driver. Windows 9x and Windows NT currently support DMA transfers by checking the Enable DMA box under Windows Device Manager. BIOS support for Ultra DMA Mode 4 (Ultra ATA/66) is also encouraged, although using software utilities you can enable the settings without BIOS support.
• A 40-pin 80-conductor cable, with a standard 40-pin 40-conductor cable Ultra ATA/66 transfer modes will be disabled and you'll basically have an Ultra ATA/33 hard drive. Ultra ATA/66 drives are backwards compatible with older host controllers, you can use them on non-Ultra ATA/33 controllers however you won't get Ultra ATA/66 performance out of the drives.  Western Digital's 13GB drive did not ship with any 80-conductor cables, and even required that Ultra ATA/66 support be turned on via their own utility available on their web site.
• Finally, you need an Ultra ATA/66 drive to complete the requirements, currently IBM (www.storage.ibm.com) and Western Digital (www.westerndigital.com) have hard drives that are Ultra ATA/66 compliant.

• AMD K6-3 400
• Prototype MVP4 board w/ 1MB L2 Cache
• 128MB PC100 SDRAM
• Western Digital Caviar AC313000 13GB Ultra ATA/66 Hard Drive
• Ultra ATAS 80-conductor cable from www.ultracable.com
• Integrated Trident Blaze 3D (8MB - MVP4 benchmarks)

All Winstone tests were run at 1024 x 768 x 16 bit color

The purpose of Ultra ATA/66 is not for a performance increase to bridge the gap between IDE and SCSI, rather it is to make room for more advancements in the storage industry, there is no point in having a slow bus being the limitation for today's hard disks. The benefits of Ultra ATA/66 will not be seen now, so don't expect to get a huge performance increase over older Ultra ATA/33 drives; however the benefits will surely come around as spin rates get faster and drives grow more efficient, wonder how long it will take before we see 10,000 RPM Ultra ATA/66 drives hit the market…